Thesis Defense - Toygar Ülgen (MSEE)

Toygar Ülgen - M.Sc. Electrical and Electronics

Asst. Prof. Göktürk POYRAZOĞLU – Advisor

Date: 18.01.2021

Time: 12:00

Location: This meeting will be held ONLINE. Please send an e-mail to gizem.bakir@ozyegin.edu.tr in order to participate in this defense.

APPLICATIONS  OF TIME-SERIES METHODS IN ELECTRICITY PRICE FORECASTING

Thesis Committee:

Asst. Prof. Göktürk Poyrazoğlu, Özyeğin University

Assoc. Prof. Cenk Demiroğlu, Özyeğin University

Asst. Prof. Emre Çelebi, Yeditepe University

Abstract:

The prediction of day-ahead electricity prices with higher accuracy is always helpful for the market players of the power exchange. Models were designed in the first place to find out the best time-series estimation method for the selected 12 European countries before and after COVID-19 using exponential smoothing family and naïve processing. Later, a classification approach is followed by 33 different features of each country to answer the question of what the best method for other countries. Also, 24 hours are estimated without any independent variables using seasonal autoregressive integrated moving average. Apart from the statistical forecasting model, long-short term memory is used for each European country. The multiple linear regression is used on electricity price forecasting. Different predictors are analyzed to reduce the mean absolute percentage error. The training data includes the dates from the day-ahead electricity market in Turkey. It is proved that the lagged electricity prices such as previous one day, one week, and lagged moving average prices play a key role in electricity price estimation. Aside from other valuable coefficients, natural gas, oil, and coal prices are tested in the forecasting model. The error rates of the fuel prices are noticeably decreased by using multiple linear regression that generates more accurate results and crucial variables influencing hourly electricity price has been determined. Different training data length is an essential part of decreasing the error proportions in the electricity price estimation. Also, it is analyzed that there is no big difference regarding the error rates if it is compared to the Regular and Dynamic Regression model in the forecast of electricity prices. Regular and Dynamic Feedforward Neural Network Methods were utilized for the model to give better error performance and more accurate results were obtained.

Bio: 

Toygar Ülgen received his B.Sc. in Electrical and Electronics Engineering from Özyeğin University, Istanbul in 2018. He is also studying for his M.Sc. degree in Electrical and Electronics Engineering at Özyeğin University. His research skills and interests are in the fields of data science, predictive analytics, time-series prediction, and machine learning and deep learning applications.